A BLIND COMPARISON OF RADIOCARBON LABS

A blind comparison of seven different radiocarbon laboratories was carried out to test reproducibility of radiocarbon age determinations. Lab A and E utilize the radiometric method, and Labs B, C, D, F and G utilize the accelerator method. Duplicate samples were sent to the laboratories without their knowledge that they were participating in the test. The results are provocative as there are significant differences among the laboratories:

Three separate macrofossils from the same sample - 18,380±100 (Lab B),17,380±130 (Lab C), and 17,820±140(Lab C).

A single Piceasp. (spruce) branch - 19,600±170 (Lab A), 19,500±170 (Lab A), 18,420±110 (Lab B), 17,290±130(Lab C), and 19,300±150 (Lab D).

A single Piceasp. (spruce) branch used as a standard by the GSC Radiocarbon Laboratory was analyzed five times resulting in an average age of 25,100±~150. The same log yielded ages of 25,320±400 (Lab A), 24,000±160 (Lab C), 25,160±100 (Lab D), 25,210±120 (Lab D), 24,710±170 (Lab E), 24,790±170 (Lab F), and 24,950±170 (Lab G).

A single Salixsp. (willow) twig - 30,400±480 (Lab B), 28,310±280 (Lab C), 33,240±230 (Lab D), and 33,140±230 (Lab D).

A single Salixsp. (willow) twig - - 24,930±940 (Lab C) and Lab D (>44,000). A second Salixsp. twig from the same sample - 33,290±380 (Lab C). A third Salixsp. twig from the same stratigraphic level - Lab D 46,500±1200 (Lab D).

The differences between laboratories generally become larger with increasing age, raising the possibility that some labs introduce a small amount of modern carbon during processing. Because chronologic control is vital for many Quaternary studies, these results are disturbing. Radiocarbon users are cautioned that different laboratories may provide significantly different ages, making correlation potentially problematic. The problem is especially acute for samples at or near the limit of radiocarbon dating. The issue of laboratory consistency aside, we also advise caution in interpreting radiocarbon ages, as they are, in reality, a probability distribution. Internal testing at the GSC Radiocarbon Laboratory demonstrates that variation between samples processed and analyzed at the same facility in some cases can exceed the combined error.